Hi, all.
I'm debugging Hadoop's source code and find an incomprehensible setting.
In hadoop-0.20.2 and hadoop-1.0.3, reducer's shuffle buffer size cannot
exceed 2048MB (i.e., Integer.MAX_VALUE). In this way, although*//*
reducer's JVM size can be set more than 2048MB (e.g.,
mapred.child.java.opts=-Xmx4000m), the heap size used for shuffle buffer
is at most "2048MB * maxInMemCopyUse (default 0.7)" not "4000MB *
maxInMemCopyUse".
I think it's not a reasonable setting for large memory machines.
The following code taken from "org.apache.hadoop.mapred.ReduceTask"
shows the concrete algorithm.
I'm wondering why maxSize is declared as "long" but set as "(int)". I
point out the corresponding code with "-->"
Thanks for any suggestion.
---------------------------------------------------------------------------------------------------------------------------
private final long maxSize;
private final long maxSingleShuffleLimit;
private long size = 0;
private Object dataAvailable = new Object();
private long fullSize = 0;
private int numPendingRequests = 0;
private int numRequiredMapOutputs = 0;
private int numClosed = 0;
private boolean closed = false;
public ShuffleRamManager(Configuration conf) throws IOException {
final float maxInMemCopyUse =
conf.getFloat("mapred.job.shuffle.input.buffer.percent", 0.70f);
if (maxInMemCopyUse > 1.0 || maxInMemCopyUse < 0.0) {
throw new IOException("mapred.job.shuffle.input.buffer.percent" +
maxInMemCopyUse);
}
// Allow unit tests to fix Runtime memory
--> maxSize = (int)(conf.getInt("mapred.job.reduce.total.mem.bytes",
--> (int)Math.min(Runtime.getRuntime().maxMemory(),
Integer.MAX_VALUE))
--> * maxInMemCopyUse);
maxSingleShuffleLimit = (long)(maxSize *
MAX_SINGLE_SHUFFLE_SEGMENT_FRACTION);
LOG.info("ShuffleRamManager: MemoryLimit=" + maxSize +
", MaxSingleShuffleLimit=" + maxSingleShuffleLimit);
}